course, the Hospitalized group has higher
viral loads than the Other group, whose viral
loads are in turn higher than those of the PAMS
group (fig. S9 and table S3). This relationship
holds across age groups (fig. S13) and also in
a fine-grained split of test centers by clinical
severity (fig. S14). Similarly, the lower first-
positive viral loads in elderly PAMS subjects
may be due to these subjects being less likely
to be tested as early because they are morelikely to be house-bound, less likely to be em-
ployed, less mobile, more cautious (therefore
disinclined to get tested with only mild symp-
toms), etc. The impact on infectiousness of
differences in viral load must be informed byJoneset al.,Science 373 , eabi5273 (2021) 9 July 2021 5 of 13
24681025 50 75 100
AgeMeanlog10viralloadSample
PAMS
Hospitalized
OtherA−3−2−1012325 50 75 100
AgeViral load difference to 50 year oldsSample
PAMS
Hospitalized
OtherB
Outcome
neg. (71)pos. (26)C0.000.250.500.751.0046810
log 10 viral loadCulture probabilitySampleHospitalized
OtherPA M S25 50 75 1000.000.250.500.751.000.000.250.500.751.000.000.250.500.751.00AgeCulture probabilityDPA M SHospitalizedOther25 50 75 100
AgeE−0.50−0.250.000.250.5025 50 75 100
AgeCulture probability difference to 50 year oldsSample
PAMS
Hospitalized
OtherFFig. 2. Estimated viral load and culture probability at time of first positive
RT-PCR test.Shaded regions denote 90% credible intervals in all panels. To
indicate change within each 90% region, shading decreases in intensity from a
narrow 50% credibility interval level to the full 90%. (A) Estimated mean viral
load in first-positive RT-PCR tests according to age and status. The stacked
histogram (right) shows the observed viral load distribution. Because the shaded
region shows the 90% credible interval for the mean, it does not include the
higher values shown in the histogram on the right. (B) Differences in estimated
first-positive viral load according to age and status. Each colored line is specific to
a particular subset of subjects (PAMS, Hospitalized, Other). Each line shows
how viral load differs by age for subjects of the corresponding status from
that of 50-year-old (rounded age) subjects of the same status. The comparison
against 50-year-olds avoids comparing any subset of the subjects against a value
(such as the overall mean) that is computed in part on the basis of that subset,
thereby partially comparing data to the same data. The mean first-positive viral loads
for 50-year-old PAMS and Hospitalized subjects are 7.2 and 6.2, respectively,
allowing relativey-axis differences to be translated to approximate viral loads.
(C) Estimation of the association between viral load and cell culture isolation success
rate based on data from our own laboratory ( 19 ) and Pereraet al.( 20 ). Viral load
differences in the log 10 range~6to~9havealargeimpactoncultureprobability,
whereas the impact is negligible for differences outside that range. The vertical lines
indicate the observed mean first-positive viral loads for different subject groups;
the horizontal lines show the corresponding expected probabilities of a positive
culture. (D) Estimated culture probability at time of first-positive RT-PCR according
to age and status, obtained by combining the results in (A) and (C). Culture
probability is calculated from posterior predictions [i.e., the posterior means shown
in (A) plus error variance]. The histogram at right shows that mean culture
probabilities calculated from observed viral loads are not well matched by credible
intervals, which do not include the most probable estimated culture probabilities.
(E) Culture probability with highest–posterior density regions, which do include
the most probable estimated culture probabilities and match the histograms in
(D) well. Theyaxis is the same as in (D). (F) Differences of estimated expected
culture probability at time of first-positive RT-PCR for age groups, with plot elements
as described for (B).RESEARCH | RESEARCH ARTICLE